Following aforementioned rationales, the aim of
this study is to investigate the readily available techniques
and provide the guideline design a properly
implementation of the flood prediction model. To
this end, MLP and RBF were chosen as our primary
schemes. Their merits were elucidated based
on GIS data recorded during the actual flooding in
2006, 2010 and 2011. Their configurations were then
adjusted accordingly to best reflect and predict the
actual events as they occurred. The resultant model
may serve as a tool in flood hazards and risks assessment.
Moreover, its deriving process could be considered
as the guidelines and precautions for adopting an
upcoming learning based flood prediction paradigm
in the future.